Towards Detecting Cocaine Use Using Smartwatches in the NIDA Clinical Trials Network (AutoSense)
|Official Title:||Towards Detecting Cocaine Use Using Smartwatches in the NIDA Clinical Trials Network|
- Ambulatory physiological and activity data [ Time Frame: 14 days ]Data yield of ambulatory physiological and activity data will be used to characterize the feasibility of using smartwatches to collect reliable interbeat interval and physical activity data in the natural field setting, sensor usage patterns (e.g., hours per day of device wear, periods of device removal), and common failure scenarios (e.g., poor sensor data quality, missing data due to removal of the device, or other factors which may affect data yield). Ambulatory physiological and activity data will be collected using passive mobile sensor data collection platforms (AutoSense chest sensors and smartwatches). This data is collected via AutoSense sensors and will utilize participants' heart beats, which will be continually streamed into the database at the Mobile Sensor Data-to-Knowledge Center of Excellence (University of Memphis). This is one assessment with one unit of measurement (heart rate) that will inform the multiple aspects of this outcome.
- Device wearability [ Time Frame: 14 days ]Device wearability (e.g., acceptability and burden) will be characterized via user questionnaires.
- Cocaine use [ Time Frame: 14 days ]Real-time self-report of cocaine use via Ecological Momentary Assessment (EMA) questionnaire prompts via the smartphone device, retrospective recall of drug use via Timeline Followback (TLFB), and urine drug assay for cocaine and other drugs of abuse will be used to adapt the computational model, so that it can be applied to the interbeat and physical activity data obtained from smartwatches. These measurements of cocaine use will be aggregated to arrive at one reported value (cocaine use episodes) to inform the model.
- Device comparison [ Time Frame: Through study completion, an average of 18 months ]Data yield will be compared between the smartwatch wrist sensor platform and the AutoSense chest sensors to characterize the conditions under which reliable interbeat interval and physical activity data can be collected with the respective devices in the natural field setting. Comparisons between overall data yield, sensor usage patterns (e.g., hours per day of device wear, periods of device removal), and common failure scenarios (e.g., poor sensor data quality, missing data due to removal of the device, or other factors which may affect data yield) will be conducted between the AutoSense chest sensors and the smartwatch wrist sensor platform to determine if differences exist between the two sensor suites. To achieve this outcome data will be aggregated into one reported value (discrepancies between the devices).
- Cocaine detection specificity [ Time Frame: Through study completion, an average of 18 months ]The degree of specificity of the cocaine detection model will be characterized relative to other stimulant use via precision and recall. To perform this comparison, rates of true positive and false positive cocaine detection events (obtained via physiological sensor data) will be compared when other stimulant drugs of abuse (e.g. amphetamines, methamphetamines) are determined to be present via urine assay for other drugs, real-time self-report of other amphetamine use via EMA, and retrospective recall of drug use via study assessments. Measures will be aggregated to arrive at one reported value, informing investigators of whether other stimulant use affects the model in the same way as cocaine.
|Study Start Date:||December 2016|
|Estimated Study Completion Date:||February 2018|
|Estimated Primary Completion Date:||February 2018 (Final data collection date for primary outcome measure)|
This study will evaluate a smartwatch device for the continuous field assessment of physiological measures, including cardiac interbeat interval and physical activity. These measures have been previously employed using wearable chest sensors to develop a model for the automatic in-the-field detection of the timing of cocaine use; computational models using physiological data of this type have been used in prior research to detect cocaine use and moment-by-moment stress using a mobile sensor suite called AutoSense. AutoSense is a chest-worn device used to collect measures of heart rate via a two-lead electrocardiograph (ECG) and physical activity via 3-axis accelerometers that can be transmitted wirelessly to an Android-based smartphone for initial processing and data storage. The adapted AutoSense protocol will incorporate smartwatches specially designed to continuously detect heart beat timings using optical photoplethysmogram (PPG) sensors rather than ECG leads.
Prior to the start of this protocol, investigators will optimize collection of cardiac interbeat interval data on the smartwatches via a preliminary ambulatory study (with Co-Investigator Ertin at the Ohio State University). The development of the smartwatch device and the initial smartwatch computational model is currently being supported separately (outside of this human subjects protocol) by the National Drug Abuse Treatment Clinical Trials Network (CTN). Investigators and research assistants at the Ohio State University will wear prototypes of the smartwatch devices and the AutoSense chest sensor during waking hours for five days to capture cardiac interbeat interval data as well as identify initial fit and usability problems with the prototype smartwatch devices and inform its subsequent refinements. This preliminary ambulatory study is a separate protocol being conducted at Ohio State University with oversight by that institution's Institutional Review Board (IRB), and thus is not considered part of the 0073-Ot human subjects protocol.
Once the preliminary study has concluded, investigators will conduct a field test during which the smartwatch and AutoSense chest sensors will be worn by 25 cocaine users for two weeks (5 participants will participate in pilot testing for two weeks each, after which the smartwatch device may undergo further refinements for improved wearability and/or data collection among the remaining participants). Outcomes of this study are to characterize the feasibility of the smartwatch device to continuously detect interbeat interval and physical activity data, and to characterize situations where data yield of sufficient quality for the application of more advanced computational models (e.g., cocaine detection) can take place in participants' natural field settings. Secondarily, data from the trial will be used to compare data yields from the two sensor suites being worn (the smartwatch devices and the AutoSense chest sensors). The data may also be useful for updating the computational models (e.g., cocaine detection) previously developed with the AutoSense chest sensors for data collected by the smartwatches. The results from this study may be used to inform future research of this type to investigate technological improvements and the situations during which using mobile sensors can unobtrusively characterize precipitants and use patterns (e.g., contextual) surrounding drug use events.
It is important to note that this study is not designed to assess the acceptability of the smartwatch among cocaine users, nor is it a study to evaluate the utility of using a smartwatch for measuring cocaine use outcomes as part of a clinical trial. Rather, investigators are intentionally recruiting participants who frequently use cocaine and compensating them to participate in this study designed to characterize the feasibility of using a smartwatch to collect reliable, continuous interbeat interval and physical activity data in the natural field setting, and to characterize under what conditions high quality data can be obtained from smartwatches. If results are promising, future research designs with larger sample sizes can explore some of these more clinically-relevant scientific questions.
Please refer to this study by its ClinicalTrials.gov identifier: NCT02915341
|Contact: Bethany M McLeman, BA||(603) 646-7086||Bethany.M.McLeman@Dartmouth.edu|
|Principal Investigator:||Lisa A. Marsch, PhD||Dartmouth College|
|Principal Investigator:||Santosh Kumar, PhD||University of Memphis|